BDA.21. Mining Complex Networks

Traditional relational data model is oftentimes inadequate when trying
to accurately capture the characteristics of large-scale and complex
phenomena. The simplicity of the relational model, which is often
considered to be one of its strongest points, becomes a serious obstacle
when trying to model complex systems, such as economy, social dynamics,
or business processes. A network data model offers more flexibility and
more expressive power.

Over the last years networks have become one of the most popular
frameworks for complex system modeling. Network perspective describes
the data from the point of view of actors and relationships between
these actors, and consequently offers more opportunities for discovering
knowledge. However, network data model introduces serious challenges
which make traditional data processing paradigms (e.g. OLAP or OLTP)
insufficient. Main challenges include:

  • the lack of well-defined structure of the data
  • high volatility and volume of data
  • the inability to use the closed world assumption
  • queries requiring excessive graph traversals

This task involves developing various algorithms for discovering
interesting patterns from complex, multi-relational networks. We are
equally interested in performing analyses on real world data, as in
developing theoretical tools for complex network mining.

Possible areas of research include:

  • statistical approach for hypothesis testing on networks
  • use of various entropies in analyzing, describing and classifying networks
  • mining frequent motives in networks
  • mining dynamic motives in networks
  • novel paradigms of network analysis
  • generative models for artificial and empirical networks
  • deep learning architectures for machine learning network problems

Main Advisor at Poznan University of Technology (PUT)
Co-advisor at